Search

CN-121995176-A - Insulator zero-value high-voltage impact detection method and system

CN121995176ACN 121995176 ACN121995176 ACN 121995176ACN-121995176-A

Abstract

The invention relates to an insulator zero-value high-voltage impact detection method and system, which belong to the technical field of insulator detection and comprise the steps of carrying out noise reduction treatment on detection signals of an insulator, carrying out feature extraction on the noise-reduced signals to obtain current-voltage features and construct defect quantification indexes, extracting insulator sub-image features by using a convolutional neural network and outputting confidence coefficients of insulator defects, extracting the temperature features of the insulator according to temperature distribution images of the insulator and constructing temperature defect identification indexes, and outputting final defect judgment results according to the defect quantification indexes, the confidence coefficients of the insulator defects and the temperature defect identification indexes. According to the multi-source information fusion method, professional noise reduction and feature extraction are carried out on the detection signals, and cross verification of the image and the temperature information is carried out, so that even if noise or shielding exists in a single channel, the single channel can be compensated by other channels, and the uncertainty of a single judgment index is reduced by the multi-source information fusion method, so that the insulator defect judgment is more stable and reliable.

Inventors

  • LI FUYIN
  • SHI ZINAN
  • YANG DAIMING
  • LIN JUN
  • GONG XIAOXU
  • LIANG HUISHI

Assignees

  • 北京西清能源科技有限公司

Dates

Publication Date
20260508
Application Date
20260309

Claims (10)

  1. 1. The zero-value high-voltage impact detection method for the insulator is characterized by comprising the following steps of: step 1, acquiring detection signals of an insulator, wherein the detection signals comprise current signals and voltage signals; step 2, carrying out noise reduction treatment on the detection signal to obtain a noise-reduced signal; step 3, extracting characteristics of the noise-reduced signal to obtain current-voltage characteristics and constructing defect quantification indexes; step 4, acquiring an insulator image, extracting the characteristics of the insulator image by using a convolutional neural network and outputting the confidence coefficient of the insulator defect; step 6, extracting temperature characteristics of the insulator according to the temperature distribution image of the insulator and constructing a temperature defect identification index; and 7, outputting a final defect judgment result according to the defect quantification index, the confidence coefficient of the insulator defect and the temperature defect identification index.
  2. 2. The method for detecting zero-value high-voltage impact of an insulator according to claim 1, wherein the step 2 of performing noise reduction processing on the detection signal to obtain a noise-reduced signal comprises: Step 2.1, performing empirical mode decomposition on the detection signal to obtain an IMF component; Step 2.2, carrying out wavelet packet transformation on the IMF component to obtain a wavelet packet decomposition coefficient; step 2.3, constructing a coefficient matrix by utilizing wavelet packet decomposition coefficients; and 2.4, carrying out singular value decomposition and noise reduction on the coefficient matrix to obtain a noise-reduced signal.
  3. 3. The method for detecting zero-value high-voltage impact of an insulator according to claim 2, wherein in step 2.2, wavelet packet decomposition coefficients are obtained by performing wavelet packet transformation on IMF components by using wavelet basis functions, and a calculation formula of the wavelet packet decomposition coefficients is as follows: ; Wherein, the Representing the jth layer kth wavelet packet decomposition coefficients, Representing the i-th IMF component, The wavelet basis functions are represented by the wavelet basis functions, Time is indicated.
  4. 4. The method for detecting zero-value high-voltage impact of an insulator according to claim 3, wherein the step 2.4 of performing singular value decomposition and noise reduction on the coefficient matrix to obtain a noise-reduced signal comprises: Step 2.4.1, carrying out singular value decomposition on the coefficient matrix, and reserving singular values larger than a threshold value to obtain a new diagonal matrix; step 2.4.2, reconstructing a wavelet packet coefficient matrix after noise reduction based on the new diagonal matrix; step 2.4.3, carrying out wavelet packet inverse transformation on the wavelet packet coefficient matrix after noise reduction to obtain IMF components after noise reduction; and 2.4.4, summing the IMF components after noise reduction to obtain a signal after noise reduction.
  5. 5. The method for detecting zero-value high-voltage impact of an insulator according to claim 4, wherein in the step 3, the current-voltage characteristics include peak value, effective value, kurtosis, form factor, impact energy, spectrum center of gravity and total harmonic distortion, and wherein the defect quantization index calculation formula is: ; Wherein, the Represents the defect quantization index, Which is indicative of the peak current value, Representing the reference peak current value, The kurtosis is indicated by the characteristic, Represents the reference kurtosis of the sample, The total harmonic distortion rate is represented by the ratio, Representing the reference total harmonic distortion rate, The energy of the impact is indicated as such, The reference impact energy is indicated as such, Which represents the peak voltage of the power supply, Representing the reference peak voltage value of the reference, A first feature weight is represented as a first feature weight, Representing the weight of the second feature, A third feature weight is represented as a third feature weight, A fourth feature weight is represented as such, A fifth feature weight is represented as a weight of the fifth feature, 。
  6. 6. The method for detecting zero-value high-voltage impact of an insulator according to claim 5, wherein in the step 6, the temperature characteristic includes a maximum temperature, an average temperature, a standard deviation of temperature, and an area ratio of hot spot areas, and the area ratio of hot spot areas is calculated by the following formula: ; Wherein, the Indicating the area ratio of the hot spot areas, Indicating the number of pixels above the temperature threshold, The length of the temperature distribution image is represented, The width of the temperature distribution image is shown.
  7. 7. The method for detecting zero-voltage impact of an insulator according to claim 6, wherein in the step 6, a temperature defect identification index calculation formula is: ; Wherein, the The temperature defect identification index is indicated to be displayed, The maximum temperature is indicated and, Indicating the maximum temperature of the reference, The reference temperature is indicated as such, The standard deviation of the reference temperature is indicated, Indicating the area ratio of the hot spot areas, 、 Representing the weights.
  8. 8. The method for detecting zero-voltage impact of an insulator according to claim 7, wherein said step 7 of outputting a final defect judgment result according to the defect quantization index, the confidence level of the insulator defect and the temperature defect identification index comprises: and outputting a final defect judgment result by adopting a weighted voting method based on the defect quantification index, the confidence coefficient of the insulator defect and the temperature defect identification index.
  9. 9. An insulator zero-value high-voltage impact detection system, comprising: The signal acquisition module is used for acquiring detection signals of the insulator, wherein the detection signals comprise current signals and voltage signals; The signal noise reduction module is used for carrying out noise reduction processing on the detection signal to obtain a noise-reduced signal; The signal characteristic extraction module is used for carrying out characteristic extraction on the noise-reduced signal to obtain current and voltage characteristics and constructing defect quantization indexes; The image defect recognition module is used for acquiring an insulator image, extracting the characteristics of the insulator image by using a convolutional neural network and outputting the confidence coefficient of the insulator defect; the temperature defect identification module is used for extracting the temperature characteristics of the insulator according to the temperature distribution image of the insulator and constructing a temperature defect identification index; the signal acquisition module is used for outputting a final defect judgment result according to the defect quantification index, the confidence coefficient of the insulator defect and the temperature defect identification index.
  10. 10. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps of a method of insulator zero value high voltage impact detection as claimed in any one of claims 1-8.

Description

Insulator zero-value high-voltage impact detection method and system Technical Field The invention relates to the technical field of insulator detection, in particular to an insulator zero-value high-voltage impact detection method and system. Background At present, the insulator is used as the most important insulating and supporting component in a power transmission line and a transformer substation, and the running state of the insulator is directly related to the safety and stability of a power grid. In the long-term operation process of the insulator, the insulator is subjected to combined action of environmental pollution, ultraviolet aging, wind blowing, rain, temperature change, lightning strike, electric arc and other electric stresses, so that the defects of internal cracks, aging of a cementing layer, umbrella skirt damage, increased creepage trace, zero-value insulator and the like are easily generated. The zero-value insulator is difficult to identify in appearance due to severe degradation or complete loss of electrical insulation performance, but the effective creepage distance and the withstand voltage of the whole string of insulators are obviously reduced, and the zero-value insulator is one of important hidden dangers for causing flashover and tripping accidents, so that the zero-value insulator has important engineering significance for effectively detecting the zero-value state of the insulator. The existing insulator defect and zero value detection method is mainly manual inspection detection, manual inspection usually relies on electric operation and maintenance personnel to visually inspect the surface of the insulator by means of a telescope, a high-definition camera and the like, and the method is sensitive to obvious defects such as mechanical damage, umbrella skirt loss and the like, but weak in the identification capability of non-appearance defects such as internal aging, zero value and the like, and the detection result is excessively dependent on personnel experience, so that the subjectivity is strong, and standardized and quantitative evaluation is difficult to realize. Disclosure of Invention In order to solve the above problems, an embodiment of the present invention is to provide a method and a system for detecting zero-value high-voltage impact of an insulator. A zero-value high-voltage impact detection method for an insulator comprises the following steps: step 1, acquiring detection signals of an insulator, wherein the detection signals comprise current signals and voltage signals; step 2, carrying out noise reduction treatment on the detection signal to obtain a noise-reduced signal; step 3, extracting characteristics of the noise-reduced signal to obtain current-voltage characteristics and constructing defect quantification indexes; step 4, acquiring an insulator image, extracting the characteristics of the insulator image by using a convolutional neural network and outputting the confidence coefficient of the insulator defect; step 6, extracting temperature characteristics of the insulator according to the temperature distribution image of the insulator and constructing a temperature defect identification index; and 7, outputting a final defect judgment result according to the defect quantification index, the confidence coefficient of the insulator defect and the temperature defect identification index. Preferably, the step 2 of performing noise reduction processing on the detection signal to obtain a noise-reduced signal includes: Step 2.1, performing empirical mode decomposition on the detection signal to obtain an IMF component; Step 2.2, carrying out wavelet packet transformation on the IMF component to obtain a wavelet packet decomposition coefficient; step 2.3, constructing a coefficient matrix by utilizing wavelet packet decomposition coefficients; and 2.4, carrying out singular value decomposition and noise reduction on the coefficient matrix to obtain a noise-reduced signal. Preferably, in step 2.2, wavelet packet transformation is performed on the IMF component by using a wavelet basis function to obtain a wavelet packet decomposition coefficient, where the wavelet packet decomposition coefficient has a calculation formula: Wherein, the Representing the jth layer kth wavelet packet decomposition coefficients,Representing the i-th IMF component,The wavelet basis functions are represented by the wavelet basis functions,Time is indicated. Preferably, the step 2.4 of performing singular value decomposition and noise reduction on the coefficient matrix to obtain a noise-reduced signal includes: Step 2.4.1, carrying out singular value decomposition on the coefficient matrix, and reserving singular values larger than a threshold value to obtain a new diagonal matrix; step 2.4.2, reconstructing a wavelet packet coefficient matrix after noise reduction based on the new diagonal matrix; step 2.4.3, carrying out wavelet packet inverse transformation on the wavelet